Introduction

Meiosis is a reductional cell division for eukaryotic sexual reproduction and generates cells that contain half of the genetic materials from the parental cells and develop into germ cells. This process involves one round of DNA replication and two rounds of nuclear division, that is, meiosis I and meiosis II. Meiosis I has a prolonged and highly complex prophase I, which has five substages with characteristic chromosomal properties: leptotene, zygotene, pachytene, diplotene and diakinesis 1, 2, 3, 4, 5. Previous studies showed that the chromosomes condense during the leptotene stage and form thin thread-like structures, then the homologs undergo pairing, synapsis and recombination from zygotene to pachytene. The homologs undergo desynapsis in diplotene, but remain associated and form highly condensed bivalents at the diakinesis stage.

Homolog paring, synapsis and recombination have very close relationships, as all three processes involve the homologs and occur during overlapping periods. Homolog pairing initiates at the leptotene-zygotene transition and progresses through zygotene 6. Homologs begin to synapse in zygotene and form synaptonemal complex (SC) at the pachytene stage. Meiotic recombination begins in late leptotene or zygotene and is completed in late pachytene. However, the nature of relationship between synapsis and recombination of homologs may differ between different organisms. In yeast, mouse and Arabidopsis thaliana, SC formation is dependent on the initiation of recombination by double-stranded DNA breaks (DSBs) 7, 8, 9, 10, 11. In Caenorhabditis elegans and Drosophila, homolog synapsis occurs earlier than recombination, and SC formation does not require DSBs, but the initiation of recombination does 12, 13.

To ensure the proper meiosis, timing and order of various meiotic events are finely controlled. Numerous studies have demonstrated that cyclins and cyclin-dependent protein kinases (CDKs) play pivotal roles in controlling major phases of the cell cycle 14, 15, 16. In yeast, besides controlling mitotic cell cycle, cyclins and CDKs also function in regulating meiosis 17, 18. In mouse, the cyclin A1 is required for normal meiosis from pachytene to diplotene 19. In Arabidopsis, a novel meiosis-specific cyclin-like protein named SDS (SOLO DANCERS) is required in homolog interaction during meiotic prophase I 20. The SDS gene is expressed in male and female meiotic cells, and sds mutant meiocytes are abnormal in synapsis, recombination and bivalent formation, resulting in abnormal chromosome distribution and defective meiotic products.

A widely accepted model for recombination, the double-stranded break repair model 21, was proposed based on the genetic and molecular studies in yeast and animal cells. According to the model, recombination is initiated by a DSB in one of the two participating molecules. Then some DSBs are converted to double-Holliday junctions, leading to the formation of recombination crossovers, whereas others are repaired via the synthesis-dependent strand annealing intermediate to form non-crossovers 22. In budding yeast, SPO11 functions in the generation of DSBs 23, 24. MSH4, MSH5 and MER3 are required for normal crossover formation 25, 26, 27, 28. MER3 is a DNA helicase and is important for extension of the DNA heteroduplex in the second end capture (SEC) intermediate 29. Recent studies showed that a MER3 homolog in Arabidopsis, named as either ROCK-N-ROLLERS (RCK) or AtMER3, is also necessary for crossover formation 30, 31. RCK is preferentially expressed in the developing anthers and T-DNA insertional rck knockout mutants show defects in homolog synapsis and crossover formation; consequently, rck late prophase I meiocytes contain lesser than five bivalents and some univalents 30.

In many organisms, the occurance of a crossover alters the probability of a nearby second crossover formation that is expected from random distribution; this phenomenon is called interference 32, 33, 34, 35, 36. Mutations in several genes, including MSH4, MSH5, MER3 and RCK/AtMER3, cause a dramatic reduction of crossovers (the residual crossovers is only about 10-15% of normal levels). Furthermore, the distributions of the remaining chiasmata in the mutants are not statistically different from predicted Poisson distributions, suggesting that the formation of the remaining crossovers is not sensitive to interference. However, the wild-type chiasma distribution is dramatically different from the Poisson distribution, demonstrating that budding yeast and Arabidopsis both possess two genetically distinct pathways for crossover formation: a major interference-sensitive pathway dependent on MSH4/5 and MER3 proteins and a minor interference-insensitive pathway that is independent of these proteins 26, 37, 38, 39, 40.

Although SDS and RCK have been shown to be important for meiosis in Arabidopsis, a member of the eudicot group of flowering plants, whether these genes represent conserved function in angiosperm is not known. The monocot rice is an excellent system for studying the molecular mechanism of meiosis 41. In recent years, a few rice meiotic genes have been reported, including PAIR1 42, PAIR2 43, OsDMC1 44, OsRad21-4 45 and OsRad21-3 46. To investigate the functional evolution of SDS and RCK, we have identified the putative rice orthologs of SDS and RCK, and studied their functions using RNAi-mediated reverse genetic approaches. Rice SDS and RCK were preferentially expressed in flowers, and suppression of either gene led to aberrant chromosome behaviors. Furthermore, crossover formation and bivalent formation were defective in RCK-deficient lines. The results indicate that rice SDS is required for normal homolog interaction and RCK is required for normal levels of crossover formation in meiosis I, revealing the conserved male meiosis mechanism in monocots and eudicots.

Results

Identification and phylogenetic analyses of putative orthologs of SDS and RCK from rice and other plants

Using the amino sequences of the Arabidopsis SDS (SOLO DANCERS) and RCK (ROCK-N-ROLLERS) proteins as queries, two rice genes, Os03g12414 and Os02g40450, encoding a putative cyclin and a putative RCK homolog, respectively, were identified by BLASTP search against the TIGR rice database. The predicted 5′ and 3′ UTRs from the Gramene database (http://www.gramene.org/Oryza_sativa_japonica/) were used to design primers for amplifying the rice SDS and RCK cDNAs, respectively, by polymerase chain reaction (PCR). The amplified cDNAs were sequenced and comparison of the cDNA and genomic sequences reveals that the predicted SDS exon/intron structure is correct, but that of RCK is incorrect. The actual RCK cDNA contains an 18-bp exon between the theoretically predicted 22nd and 23rd exons.

Similar to the Arabidopsis SDS protein, the predicted rice SDS protein contains two conserved cyclin domains: cyclin_N (spanning amino acids 239-375) and cyclin_C (amino acids 377-462) (Figure 1A, top panel), suggesting that rice SDS might have the activity to regulate CDKs. Sequence alignment showed that rice SDS has 49% identity and 64% similarity with SDS of Arabidopsis.

Figure 1
figure 1

Structural organization and phylogenetic analyses of putative orthologs of rice SDS and RCK. (A) Structural organization of rice SDS and RCK. Positions of domains cyclin_N, cyclin_C, DEXDc, HELICc and SEC63 are indicated. (B) A neighbor-joining phylogenetic tree for SDS and its homologs in different organism, with bootstrap values higher than 50% shown for each clade. Bootstrap values in parentheses are from the tree generated using Maximum Likelihood method for the same sequence alignment. Except indicated, those homologous are with following accession numbers: S. moellendorffii (jgi|Selmo1|231762|fgenesh1 pm.); P. patens-1 (jgi|Phypa1_1|186054|estExt gwp); P. patens-2 (jgi|Phypa1_1|32594|gw1.147.60.); Z. mays (AC205249.1_FGP020); P. trichocarpa (eugene3.00100975) and V. vinifera (CAN78702). (C) A neighbor-joining phylogenetic tree of RCK and its homologs in different organisms. The bootstrap values in parentheses are given for Maximum Likelihood analysis. Except those indicated, those homologous are with following accession numbers: S. bicolor (jgi|Sorbi1|4812687|e gw1.4.145); Z. mays (AC188828.2_FG028); P. patens (jgi|Phypa1_1|122688|e gw1.41.1) and S. moellendorffii (Selmo1|423377|fgenesh2 pg.C sc).

Full size image

Sequence analysis showed that rice RCK has several conserved domains, including DEXDc (amino acid positions 33-224), HELICc (amino acids 330-442) and SEC63 (amino acid positions 538-852) (Figure 1A, bottom panel). DEXDc is found in members of a DEAD-like superfamily involved in ATP-dependent RNA or DNA unwinding 47. The HELICc domain is located at C-terminus of proteins that belong to a helicase superfamily, and is associated with DEXDc-, DEAD- and DEAH-box proteins 48. The SEC63 domain was first found in the budding yeast SEC63 protein 49, 50. These domains are also found in the yeast MER3 and Arabidopsis RCK proteins and the presence of these domains suggests that RCK possibly has a helicase activity. The protein structures of rice SDS and RCK suggest that these two proteins may have conserved functions in rice. Further sequence comparison indicated that rice RCK has 73% identity and 87% similarity with Arabidopsis RCK.

To test whether SDS and RCK are conserved, we searched for sequences of additional putative SDS and RCK homologs and other related proteins. Putative SDS homologs from poplar (Populus trichocarpa), grape (Vitis vinifera), maize (Zea mays) and sorghum (Sorghum bicolor) were identified, as well as representative genes for cyclin A and B. Alignment and phylogenetic analysis using neighbor-joining and maximum likelihood methods yielded trees with the same topology (Figure 1B). The six SDS orthologs form a clade with high bootstrap support separate from the clades for cyclins A and B, suggesting that SDS genes are derived from a single gene in the ancestor of flowering plants. This is consistent with the previous finding that the Arabidopsis and rice SDS genes form a distinct clade among plant cyclin genes 51. Because the cyclin A clade also contains animal genes, this further suggests that the SDS clade was formed prior to the divergence of animal and plants. We have also retrieved sequences of putative RCK homologs from several plants and representative non-plant species and constructed a phylogenetic tree (Figure 1C). The topology of the RCK is in agreement with the current understanding of the eukaryotic species relationship, indicating that the plant RCK genes are likely orthologs among themselves.

SDS and RCK are both preferentially expressed in flowers

Sequence and phylogenetic analyses indicate that SDS and RCK are conserved in angiosperms, suggesting that they might have similar functions in flowering plants. To obtain clues about their functions, the expression of SDS and RCK in rice was examined by real-time quantitative RT-PCR using total RNAs extracted from vegetative and reproductive tissues. Floral stages were determined according to floral length and cytological observations, essentially as described previously 46. Flowers with lengths of ≤ 0.8 mm were pooled as F1 (stamen and carpel primordial formation), flowers of 0.8-2 mm in length as F2 (pollen mother cell formation), those 2-4 mm long as F3 (male meiosis) and flowers of 4-6 mm in length as F4 (microspore formed). Flowers of 6 mm or longer, but before the heading stage and with anthers that were turning yellow were pooled as F5 (bicellular pollen). In the panicles (inflorescences) after heading, flowers close to anthesis were pooled as F6 (tricellular pollen stage). Quantitative real-time PCR analysis revealed that SDS was expressed preferentially in flowers and stems, with relatively weak expression in roots and leaves (Figure 2A, left panel). The level of SDS transcript peaked at F1-F3 and decreased thereafter. Similarly, RCK was expressed preferentially at F1-F3 as well (Figure 2A, right panel). These results suggested that both SDS and RCK were expressed at early stages of meiocyte development and during meiosis.

Figure 2
figure 2

Expression of SDS and RCK. (A) Quantitative real-time PCR analysis reveals the expression of SDS (left panel) and RCK (right panel) in various tissues and developing flowers, including root (R), stem (S), leaves (L), stamen and carpel primordial formation (F1), pollen mother cell formation (F2), male meiosis (F3), uninucleate microspore (F4), bicellular pollen (F5) and tricellular pollen (F6). Relative mRNA levels were determined by normalizing the PCR threshold cycle number of each gene with that of the ACTIN1 reference gene. The results are presented as an average of three independent experiments. (B) Promoter-reporter gene fusion studies indicated that rice SDS is transcribed in the shoot apex of stem (2), panicle (4) and anther (5); and no expression is seen in 4-day seedling (1), mature leaf (3) and grains in different development stages (6). Bar = 1 mm. (C) Histochemical analysis revealed that rice RCK is transcribed in stem node (2), mature leaf (3), flowers (4), flower with lemma and palea removed (5), panicle (6, highlighted in 7, 8) and grains in different development stages (9). Bar=100 μm (8) and 1 mm (1-7, 9).

Full size image

To investigate the expression patterns of SDS and RCK in detail, reporter gene fusion studies were performed with the SDS or RCK promoters and the GUS gene coding for β-glucuronidase (GUS). More than 20 independent transformants were obtained and analysis of the GUS activities in 10 transgenic lines revealed that SDS was expressed in the shoot apex (Figure 2B, 2) and flowers (Figure 2B, 4), preferentially in anthers (Figure 2B, 5). On the other hand, SDS expression was not detected in seedlings (Figure 2B 1), leaves (Figure 2B, 3) and grains at different developmental stages (Figure 2B, 6). RCK was expressed in the node of stem (Figure 2C, 2), leaves (Figure 2C, 3), pistils of the mature flower (Figure 2C, 4, 5), panicles (Figure 2C, 6) and panicles before meiosis stage (Figure 2C, 7), anthers and pistils of flowers before meiosis stage (Figure 2C, 8), and developing seeds (Figure 2C, 9).

Suppression of SDS results in defective pollen and reduced fertility

To characterize the in vivo function of SDS, we generated a gene-specific p35S SDS-RNAi construct using a 535-bp fragment of SDS cDNA (nt 430-964) that shares no similarity to any other sequences in the rice genome. A total of 30 independent transformants were identified by hygromycin resistance and the presence of the T-DNA in the rice genome was verified by PCR with primers localized to the spacer and the inserted cDNA sequences of vector p35S SDS-RNAi. Further analysis using real-time RT-PCR indicated that the SDS expression was suppressed and lower than that in WT (Figure 3A). The SDS-RNAi lines developed normally during the vegetative stage, but seed production was severely disrupted as compared to WT plants. The mean seed setting rate (number of seeds as a percentage of the total number of flowers) at maturity of p35S SDS-RNAi lines was greatly reduced (23.0%, 61.5%, 2.1% and 46.8% of lines L4, L8, L14 and L18, respectively), compared to that of WT (93.8%) (Figure 3D), indicating that suppression of SDS in the RNAi lines results in a sterility phenotype.

Figure 3
figure 3

Suppressed expression of SDS resulted in reduced number and viability of pollen grains, and delayed seed setting rate. (A) Quantitative real-time PCR analysis of SDS mRNA levels in WT and independent transgenic lines (L4, 8, 14, 18) by RNAi strategy. Total RNAs were extracted from flowers. mRNA levels are normalized relatively to the ACTIN1 level. Results are presented as an average of three independent experiments. (B) Viability of mature pollen grains in WT (left) and representative transgenic line (p35S SDS-RNAi-L14, middle), examined with iodium potassium iodide solution (left panel). The relative number of pollen grains is counted using a hematocyte arithmometer, with that of WT as control (right). Statistical analysis using one-tailed Student's t-test indicated the significant difference (*P < 0.05). Bar = 200 μm. (C) Pollen viability analysis of WT (left) and representative transgenic line p35S SDS-RNAi-L14 (middle) with FDA staining. The relative viability is calculated and shown in percentage (right). Statistical analysis using one-tailed Student's t-test indicated the significant differences (**P < 0.01). Bar = 200 μm. (D) Growth of plant (left), panicle (middle) and seed setting rate (right) of WT and transgenic line p35S SDS-RNAi-L14 (T1 generation). Statistical analysis using one-tailed Student's t-test indicated the significant differences (**P < 0.01).

Full size image

As pollen development is important for the successful fertilization of sexual plants, the pollen development of rice SDS-RNAi lines (L4, L8, L14 and L18) in the T1 generation was compared with that of wild type. Mature anthers from the deficient lines did not differ obviously in morphology from those of wild type, but in an I2-KI staining assay, pollen grains from the wild type were round and had a uniform size, with a dark blue-black reaction (Figure 3B, left panel); in contrast, those of the SDS-RNAi lines were variable in size and shape (Figure 3B, middle panel). We also estimated the pollen amount per flower in these lines and found that the SDS-RNAi lines contained decreased amounts of pollen grains than wild type. Each flower of the SDS-RNAi lines L4, L8, L14, L18 had 73.3%, 77.6%, 68.6% and 73.8% of the wild-type amount of pollen grains on average, respectively (Figure 3B, right panel).

Further examination on the pollen viability using fluorescein diacetate (FDA) showed that the SDS-RNAi lines had significantly reduced pollen viability. Compared to 91.2% viability of WT pollen grains (n = 4 889), those of SDS-RNAi lines had 61% (L4, n = 2 344), 76.4% (L8, n = 2 667), 35.1% (L14, n = 2 187) and 75.4% (L18, n = 4 188) viable pollens, respectively (Figure 3C), consistent with the levels of suppression of SDS expression, indicating that pollen viability is disrupted under SDS suppression.

Suppression of RCK results in reduced pollen viability and reduced fertility

The physiological function of rice RCK was studied by a similar RNAi approach. Briefly, a rice RCK RNAi vector (p35S RCK-RNAi) was constructed with a 514-bp cDNA fragment of RCK (nt 2 409-2 922), which shares no similarity to other sequences in the rice genome. After transformation, 40 independent T0 transformants were obtained and most of them were confirmed to carry the T-DNA by PCR analysis. Further expression analysis on the 10 selected plants by quantitative real-time RT-PCR analysis confirmed the suppressed expression of endogenous RCK in several independent lines (L9, L12, L13 and L19), especially L12 (Figure 4A).

Figure 4
figure 4

Suppressed expression of RCK resulted in an altered viability of pollen grains, and seed setting rate. (A) Quantitative real-time PCR analysis of RCK mRNA levels in WT and independent transgenic lines (L9, 12, 13, 19) by RNAi strategy. Total RNAs were extracted from flowers. mRNA levels are normalized relatively to the ACTIN1 level. The results are presented as an average of three independent experiments. (B) Viability of mature pollen grains in WT (left) and representative transgenic line (p35S RCK-RNAi-L12, right), examined with iodium potassium iodide solution (left panel). Bar = 200 μm. (C) Pollen viability analysis of WT (left) and representative transgenic line p35S RCK-RNAi-L12 (middle) with FDA staining. The relative viability is calculated and shown in percentage (right). Statistical analysis using one-tailed Student's t-test indicated the significant differences (**P < 0.01). Bar = 200 μm. (D) Growth of plant (left), panicle (middle) and seed setting rate (right) of WT and transgenic line p35S RCK-RNAi-L12 (T1 generation). Statistical analysis using one-tailed Student's t-test indicated the significant differences (**P < 0.01).

Full size image

Similar to the SDS-RNAi lines, RCK-RNAi lines grew and developed normally during their vegetative stage, but seed production was severely disrupted. Calculation on the mean seed setting rate at seed maturity showed that, compared to WT plants (seed setting rate of 88.2%), RCK suppression resulted in the rate of 26.7%, 4.81%, 12.1% and 36.8% (lines L9, L12, L13 and L19), respectively (Figure 4D), suggesting that knockdown of rice RCK in RNAi lines is related to their sterility.

Observation on the pollen development and fertility showed that RCK suppression did not result in altered morphology of mature anthers; however, in an I2-KI staining assay, pollen grains of the RCK-RNAi lines were variable in size and shape. Pollen grains from the WT were round and had a uniform size, with a dark blue-black staining (Figure 4B, left panel). In contrast, the number of pollen grains that were stained (viable) was greatly reduced in the RCK-RNAi lines and some grains were empty, shrunken and stained brown (Figure 4B, right panel). The pollen amount per flower was similar to that of WT, unlike the SDS-RNAi lines.

The FDA staining assay indicated that the pollen viability of the RCK-RNAi lines was significantly reduced. Compared to the 91.4% pollen viability of WT (n = 2 427), those of RCK-RNAi lines were only 68.7% (L9, n = 4 023), 45.8% (L12, n = 2 057), 47.9% (L13, n = 4 321) and 75.4% (L19, n = 3 179), respectively (Figure 4C), indicating that pollen viability was disrupted in RCK RNAi lines.

Suppression of SDS and RCK results in defective male meiosis I

To evaluate whether the abnormal pollen development resulted from defective meiosis, we performed detailed cytological analysis of meiosis in male meiocytes from the SDS-RNAi (L4, L8, L14 and L18) and RCK-RNAi lines (L9, L12, L13 and L19). In WT, leptotene chromosomes appeared as thin threads that looped out of dense synizetic knots (Figure 5A). Homologous chromosomes began to associate side by side at zygotene (Figure 5B), and fully synapsed into thick threads at pachytene (Figure 5C). After synapsis was resolved at diplotene, except at chiasmata (Figure 5D), 12 bivalents became tightly condensed at diakinesis (Figure 5E) and were arranged at the equatorial plate at metaphase I (Figure 5F). At anaphase I, homologs separated (Figure 5G) and moved towards opposite poles of the spindle, forming two clusters at telophase I (Figure 5H).

Figure 5
figure 5

Male meiosis I in WT and transgenic lines with suppressed expressions of SDS or RCK. The DAPI-stained chromosomes are shown. Prophase I at leptotene, zygotene, pachytene and diplotene, are similar in WT (A-D), p35S SDS-RNAi lines (I-L) and p35S RCK-RNAi lines (Q-T), while the diakenesis of prophase I, metaphase I, anaphase I and telophase I are significantly altered. At diakenesis of prophase I, 12 attached pairs of condensed homologs are formed in WT (E); while 24 stained bodies can be observed in p35S SDS-RNAi lines, indicating that the condensed homologs formed univalents (M); 16 stained bodies can be observed in p35S RCK-RNAi lines, indicating a mixture of bivalents and univalents (U). At metaphase I, the chromosomes align at the equator in WT (F); while those that did not align on equator in p35S SDS-RNAi lines (arrowheads, N) or did not all align at the equator in p35S RCK-RNAi lines (arrowheads highlight that some chromosomes are away from the equator, V). At anaphase I, the homologs are separating in WT (G); while the chromosomes are elongated similarly to those at metaphase I in p35S SDS-RNAi lines, suggesting that they might be pulled by the spindle as normal homologs are at anaphase I (arrowhead highlights the lagging chromosomes, O); or some chromosomes are with delayed elongation (arrowheads) in p35S RCK-RNAi lines (W). At telophase I, the homologs separate and form two clusters in WT (L); while some chromosomes more distant from the center are decondensed in p35S SDS-RNAi lines, resembling those in normal telophase I (arrowhead and arrow highlight that the chromosomes still stayed at the center and chromosome bridges, respectively, P); in p35S RCK-RNAi lines, most chromosomes are decondensed normally, but some are delayed (arrowheads, X). Bar = 10 μm.

Full size image

In both SDS and RCK RNAi lines, chromosomal patterns similar to the WT leptotene through pachytene stages were observed (Figure 5I-5K, 5Q-5S). Although it was difficult to determine whether diplotenes in SDS- and RCK-deficient lines were normal (Figure 5L and 5T), it was obvious that, by diakinesis, the homologs were not all attached in SDS-RNAi cells with 24 univalents (Figure 5M). In RCK-RNAi cells, there were more than 12 distinguishable chromosomes, indicating the presence of univalents (Figure 5U). Furthermore, comparison of WT and transgenic meiocytes at late stages of meiosis I revealed additional defects, which may be a result of the defective prophase I (Figure 5N-5P and 5V-5X). In p35S SDS-RNAi-L14 and p35S RCK-RNAi-L12 cells at metaphase I, chromosomes often did not align at the equatorial plane (Figure 5N and 5V); some of them were quite far from the equator (arrowheads in Figure 5N and 5V). The anaphase I cells showed more than two clusters of chromosomes and several lagging chromosomes (arrowheads in Figure 5O and 5W). Then at telophase I, in p35S SDS-RNAi-L14 cells, some chromosomes still stayed at the center (arrowhead in Figure 5P) and chromosome bridges were observed (arrow in Figure 5P); in p35S RCK-RNAi-L12 cells, some chromosomes were delayed in moving to the poles (arrowheads in Figure 5X).

Altered meiocyte distributions under SDS or RCK suppression

Male meiosis in rice is slightly asynchronous and a population of male meiotic cells from a single flower can cover a few adjacent meiotic stages or substages of the long prophase I 41. We noticed that p35S SDS-RNAi and p35S RCK-RNAi cells with chromosome images similar to that in Figure 5K and 5S (pachytene) were very infrequent, suggesting defects in synapsis. Then, we examined a large number of meiocytes from wild type, p35S SDS-RNAi-L14 and p35S RCK-RNAi-L12 to analyze the meiocyte distribution, similar to the examination of the Arabidopsis sds mutant 20. The results showed that among the wild-type prophase I cells (1954 cells), nearly 31% were pachytene cells and relatively few were leptotene cells (Figure 6A). In contrast, the p35S SDS-RNAi-L14 showed a dramatically different distribution of prophase I substages (2 138 cells); the number of cells at the zygotene was much higher than normal, whereas cells at the pachytene were rare. The pattern of p35S SDS-RNAi-L14 distribution of prophase I stages supports the idea that synapsis is defective in SDS-deficient transgenic lines. Similarly, the distribution patterns of wild-type (3 232 cells) and p35S RCK-RNAi-L12 (3 842 cells) meiosis I meiocytes indicate that p35S RCK-RNAi-L12 also had increased frequencies of cells at the leptotene and zygotene stages, and reduced frequencies of cells at diakenesis, metaphase I, anaphase I and telophase I (Figure 6B).

Figure 6
figure 6

Distribution of meiosis I meiocytes of WT, p35S SDS-RNAi and p35S RCK-RNAi lines. Chromosome spreads are examined as shown in Figure 5. Lep (leptotene), Zyg (zygotene), Pac (pachytene), Dip (diplotene), Dia (diakenesis of prophase I), Met (metaphase I), Ana (anaphase I) and Tel (telophase I). (A) Distribution of prophase I cells in WT (1 954 cells) and the p35S SDS-RNAi-L14 line (2 138 cells). (B) Distribution of meiosis I meiocytes of WT (3 232 cells) and p35S RCK-RNAi-L12 line (3 842 cells).

Full size image

Suppression of RCK results in decreased numbers of chiasmata and bivalents

The fact that some bivalents were observed in RCK-RNAi transgenic meiocytes indicates that some crossovers were formed. To examine this aspect of the meiotic phenotype, we counted the number of chiasmata. At early diakinesis, meiocytes of the RCK-deficient lines exhibited a reduction in chiasma number compared with wild-type meiocytes (Figure 7A). A sample of 140 p35S RCK-RNAi-L12 cells showed an average of 13.99 chiasmata per cell compared with an average of 23.90 chiasmata per cell in the wild-type meiocytes. Several univalents were observed in the p35S RCK-RNAi-L12 cells (arrows in Figure 7A, right panel) in contrast to 12 bivalents during the comparable stage in the wild-type cells (Figure 7A, left panel). Among 140 p35S RCK-RNAi-L12 meiocytes at late prophase I to metaphase I, no meiocytes had 12 bivalents; 0.7, 0.7, 3.6, 7.9, 6.4, 8.6, 27.9, 19.3, 12.1, 7.1, 5.7% were observed with 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11 bivalents, respectively, with an average of 7.21 bivalents per cell (Figure 7C).

Figure 7
figure 7

Distribution of chiasmata and bivalents of WT and p35S RCK-RNAi line. (A) Early diakinesis of WT (left) and p35S RCK-RNAi-L12 (right). Reduced chiamasta can be observed. Arrows highlight some univalents. Bar = 10 μm. (B) Frequency of chiasma number per meiocyte in WT (top panel, χ2(20) = 364.65, P < 0.005, n = 100) and p35S RCK-RNAi-L12 line (bottom panel, χ2(17) = 32.73, 0.01 < P < 0.025, n = 140). PD, predicted Poisson distribution; OD, observed distribution. (C) The frequency of bivalent number per meiocyte in WT and p35S RCK-RNAi-L12 lines.

Full size image

We analyzed the distribution of residual chiasmata present in p35S RCK-RNAi-L12 and compared them with that of the wild type (Figure 7B). The chiasma distribution for wild-type cells (n = 100) deviated significantly from the Poisson prediction (Figure 7B, top panel; χ2(20) = 364.65, P < 0.005), whereas the chiasma distribution in p35S RCK-RNAi-L12 cells (n = 140) was much closer to the predicted Poisson distribution (Figure 7B, bottom panel; χ2(17) = 32.73, 0.01 < P < 0.025). The results demonstrate that RCK deficiency likely causes severe defects in the interference-sensitive pathway of chiasma formation.

Discussion

Evolution of SDS and RCK in flowering plants

We have identified SDS and RCK homologs from several flowering plants, and phylogenetic analyses indicate that SDS and its homologs form a well-supported clade separate from other cyclin genes, such as cyclin A and cyclin B genes. This phylogenetic relationship indicates that the SDS homologs from flowering plants are likely true orthologs. For RCK, homologs were identified in several plants, as well as animals and fungi. The phylogenetic relationship of RCK homologs is in agreement with the accepted relationship of the corresponding organisms, strongly suggesting that these genes are also orthologous. In addition, only one copy of SDS and one copy of RCK were found in Arabidopsis, poplar, and rice, plant species whose whole genome has been sequenced. It has been proposed that during the evolution of these species, there have been one or more genome-wide duplication(s). Therefore, additional copies of SDS must have been lost after the duplication(s), unlike the cyclin A and cyclin B genes, indicating that SDS has a different pattern of evolution from those of other types of cyclin genes, which have expanded in flowering plants. The RCK gene has also been maintained as single copy during eukaryotic evolution. The loss of duplicated SDS and RCK genes in flowering plants suggests that additional copies are not beneficial, or even possibly deleterious to plants.

Roles of SDS and RCK in rice male meiosis

The observation that SDS and RCK are maintained as single copy genes in angiosperms suggests that they have conserved functions. Our studies using RNAi rice plants indicated that SDS and RCK are needed for normal rice meiosis and male fertility. In SDS-RNAi meiocytes, chromosome condensation during prophase I was normal, but there were many more zygotene cells and very few pachytene cells, suggesting that normal synapsis of homologs cannot be achieved under SDS deficiency, although a defect in pairing is also possible. Together with the observations of individual univalents at late prophase I, we conclude that the rice SDS gene is required for homolog interaction and bivalent formation, similar to the Arabidopsis SDS function 20. RCK-RNAi meiocytes also exhibited defects in meiotic crossover and bivalent formation. Compared to WT, the number of chiasmata in RCK-RNAi cells was dramatically reduced and the distribution of them was much closer to the Poisson distribution of random events, though the statistical analysis still showed a significant difference between the distribution of RCK-RNAi-L12 and Poisson distribution (Figure 7B, bottom panel). The relative few and close-to-randomly distributed crossovers in the RCK-RNAi plants suggest that rice, like Arabidopsis 3, also has two pathways for crossover formation: a major interference-sensitive RCK-dependent pathway and a minor pathway that is RCK-independent and interference insensitive. However, because the crossovers in the RCK-RNAi plants were not completely randomly distributed, the existence of a minor interference-insensitive pathway in rice, as that in Arabidopsis, needs further investigation. Phenotype of RCK-RNAi cells indicated that they were defective in the interference-sensitive pathway of meiotic recombination, similar to the phenotype of the Arabidopsis rck mutants 30. Therefore, the functions of SDS and RCK are largely conserved in Arabidopsis and rice, which represent eudicots and monocots, respectively.

The relationship between recombination and synapsis is one of the most actively investigated aspects of meiosis. In yeast and Arabidopsis, synapsis and recombination are interdependent and closely coupled. In yeast, an important SC component HOP1, can preferentially bind to yeast DNA and promote crossover formation in meiotic recombination 52, 53. Genetic studies strongly support the idea that SC is dependent on the initiation of recombination by DSBs. A mutation in the Arabidopsis ASY1 gene encoding an HOP1 homolog also causes a reduction in chiasmata number 54, 55. In addition, Arabidopsis sds mutant also shows greatly reduced frequency of meiotic recombination 20, and some other mutants defective in meiotic recombination including atspo11-1, atrand51, atxrcc3 and rck 8, 30, 56, 57 are unable to achieve normal SC formation, indicating that normal levels of recombination are needed for proper SC formation. Because SDS could potentially regulate gene expression indirectly, we tested whether rice SDS can affect RCK expression and found that the level of RCK transcript was not influenced when SDS expression was suppressed (Figure 8), suggesting that, though SDS and RCK function in two closely coupled processes of synapsis and recombination, they are not regulated by each other in rice.

Figure 8
figure 8

Evolutionary relationship between SDS and RCK. Real-time RCR analysis on the RCK transcripts in WT and p35S SDS-RNAi lines (L4, 8, 14, 18). Total RNAs were extracted from flowers, and results are presented as an average of three independent experiments.

Full size image

Conserved male meiosis in monocots and eudicots

In addition to the conserved sequences and functions of rice SDS and RCK shown in this study, a small number of rice meiosis genes have been studied including PAIR2 43, OsDMC1 44, OsRad21-4 45 and their functions in rice male meiosis are conserved compared to their correspondent homologous genes ASY1 58, AtDMC1 59, 60, SYN1 61 in Arabidopsis. In Arabidopsis, the ASY1 protein is localized to the axial/lateral elements and an asy1 mutant is asynaptic and unable to form SC in both male and female meiocytes, indicating its critical role in SC formation 54, 58, 62. A mutation in PAIR2, the rice homolog of ASY1, causes defects in homolog alignment at pachytene and the formation of univalents instead of bivalents at diakinesis 63. The PAIR2 proteins associate with axial elements (AEs) at leptotene and zygotene, and are removed from the AEs of arm regions when homologous chromosomes are synapsed 43. The Arabidopsis dmc1 mutant forms univalents instead of bivalents at meiotic prophase I and the recombination frequency is greatly reduced 59. Similarly, knock down of rice DMC1 leads to defects in bivalent formation and subsequent unequal chromosome segregation 44. In Arabidopsis, sister chromatid cohesion is abnormal and chromosome condensation is also affected in the syn1 mutant, and meiocytes of the mutant produce chromosomal fragments 64, 65. The male meiocytes in OsRAD21-4-deficient lines showed over-condensation of chromosomes, precocious segregation of homologs and chromosome fragmentation. Fluorescence in situ hybridization revealed that OsRAD21-4-deficient lines were defective in homologous pairing and cohesion at sister chromatid arms 45. These results strongly support a conserved molecular mechanism for male meiosis in monocots and eudicots.

Possible modified functions of SDS and RCK in Oryza sativa

It is worth noting that some cytological characteristics in rice meiosis are different from that of Arabidopsis. The rice meiosis I is followed by the formation of a cell plate 41, instead of an organelle band, as observed in Arabidopsis, between the two daughter cells, indicating the differential aspects in meiosis of rice and Arabidopsis. Indeed, expression pattern analysis of the rice SDS and RCK genes demonstrated that although both of them were expressed preferentially in flowers, they were transcribed in other tissues as well. SDS was highly expressed in the shoot apex, suggesting that it might function in mitosis, which is different from the Arabidopsis SDS gene that was only detectable in flowers containing meiotic cells and not in vegetative organs or other stages of reproductive structure 20. Additionally, rice SDS expression in flowers was only detected in anther but not in pistils, which might be different from the Arabidopsis SDS, which functions in both female and male meiosis. Besides flowers, rice RCK was also expressed in stems, leaves and seeds. Specifically, its expression was changed during floral development. In the flower before meiosis, RCK expression was detected in both anthers and ovaries, while after meiosis stage, RCK was mainly expressed in the ovaries. When the p35S RCK-RNAi-L12 lines were pollinated with wild-type pollen, the mean seed setting rate was only 12.4%, which is much lower when compared to the mean seed setting rate of wild-type lines (53.9%), indicating that suppression of rice RCK also reduced the female fertility. These results suggest that rice RCK may have a modified function in male and female meiosis. In p35S RCK-RNAi meiocytes, there were many zygotene cells and very few pachytene cells, similar to SDS-RNAi meiocytes; while the Arabidopsis rck mutants had an increased frequency of cells at the diakenesis stage 30. Therefore, besides the conserved functions in chiasma and bivalent formation, rice RCK may be required for other events of homolog interactions during prophase I such as synapsis. Further analysis is needed to determine the functional differences of SDS and RCK in rice.

Materials and Methods

Plant materials

Rice plants (Oryza sativa, japonica cultivar Zhonghua 11) were grown in a phytotron with a light/dark cycle of 12 h at 28 °C/12 h at 22 °C. Flowers in different stages were collected. Leaves and stems were collected from 6-week-old plants. Roots were collected from 1-week-old seedlings germinated in water.

Isolation of rice SDS and RCK

The cDNA sequences of SDS and RCK were amplified by RT-PCR using the first-strand cDNA synthesized from rice flower RNAs as a template, with specific primers SDSF1 (5′-GGC CCG TAG CAT CAC AAA TTC AC-3′) and SDSR1 (5′-CGG CAT ATC ACC TGG GTT TGT TC-3′), and RCKF1 (5′-CGA GAC AGT AGC GCT AGC ACG AAC A-3′) and RCKR1 (5′-TGC TGC ACA AGA TCG CCA CTG-3′). The amplified cDNA fragments were gel-purified, subcloned into the pGEM-T vector (Promega) and sequenced.

Promoter-GUS fusion studies

For further analysis of the expression pattern of SDS and RCK, the 1.9-kb promoter region of rice SDS was PCR amplified with primers SDSP1 (5′-TGG CGT TGG CTA AGA AGG TTG CTG-3′) and SDSP2 (5′-TGG CGT GCG GCG CTG TGT G-3′); the 2.5-kb promoter region of RCK was amplified with primers RCKP1 (5′-CGC GCA TGA ACA CGA TCC AAG-3′) and RCKP2 (5′-CGC GTG CGG TCC CTG CAG-3′), using rice genomic DNA as template. The resultant DNA fragments were subcloned into vector pCambia1300+pBI101 66, resulting in the fusions with GUS reporter gene. The obtained constructs were introduced into Agrobacterium tumefaciens (strain EHA105) and transformed into the rice genome using immature embryo as materials 67. Positive transgenic plants were obtained through selection by hygromycin resistance and PCR amplification. GUS activities were detected histochemically according to Jefferson et al. 68. Photography was performed using a Nikon microscope with a digital camera.

RNAi constructs

A 478-bp intron amplified from rice genome with primers InF (5′-GTC TCTAGA GTA AGT TAC AAA CCT TTT TGT AC-3′, XbaI site underlined) and InR (5′-GCC GTCGAC CTG AAA ATC TCG AAA CAG CCG TGT C-3′, SalI site underlined) was ligated into a pSK vector after digestion with XbaI and SalI, resulting in a vector pSK-Int. Then, two 535-bp fragments specific to SDS cDNA (nt 430-964) were amplified by PCR with primers SDSF2 (5′-AGCTCGAGT CCT ACC TCG CCT GCC CCG-3′, XhoI site underlined) and SDSR2 (5′-TAGTCGACG CCC TAC CCC ATG AAC ACG GTC-3′, SalI site underlined), and primers SDSF3 (5′-TAGGATCCT CCT ACC TCG CCT GCC CCG-3′, BamHI site underlined) and SDSR3 (5′-TATCTAGAG CCC TAC CCC ATG AAC ACG GTC-3′, XbaI site underlined). After digestion with corresponding restriction enzymes, the two fragments were subcloned into vector pSK-Int to generate the construct pSK-SDSi. Finally, the fragment in pSK-Int containing the 478-bp intron flanked with two 535-bp opposite fragments was digested with XhoI and BamHI and subcloned into pCAMBIA1300S, resulting the RNAi construct, p35S SDS-RNAi.

In a similar strategy, two 514-bp fragments specific to RCK cDNA (nt 2 409-2 922) were amplified by PCR with primers RCKF2 (5′-AGCTCGAGT GCT CCT TGC GAA ATG CTT ACA C-3′, XhoI site underlined) and RCKR2 (5′-TAGTCGACT CCC TGC TGA CCA CAT GCT TCT TG-3′, SalI site underlined), and primers RCKF3 (5′-TAGAGCTC T GCT CCT TGC GAA ATG CTT ACA C-3′, SacI site underlined) and RCKR3 (5′-TATCTAGAT CCC TGC TGA CCA CAT GCT TCT TG-3′, XbaI site underlined). The obtained DNA fragments were used to generate the RNAi vector of RCK, p35S RCK-RNAi. The resultant constructs were transformed into rice genome by A. tumefaciens-mediated transformation.

Phenotypic analysis

Fluorescein diacetate and I2-KI were used to determine the viability of pollen grains. Mature flowers before opening were harvested and picked anthers were crushed into a fine powder and stained with 10 μl I2-KI solution (1 g KI and 0.5 g I2 dissolved in 100 ml distilled water) for a couple of minutes 69 and observed under a microscope (Leica DMR, Germany). Pollen grains that were round in shape and stained dark blue-black were regarded as viable or living pollen. The sterile or dead pollens were stained brown.

Anthers were treated with FDA solution (10 mg FDA dissolved in 1 ml acetone as stock, diluted with 7% sucrose to a concentration of 100 μg/ml for use) for 5 min, and then photographed under ultraviolet light (Leica DMR, Germany) 70. Pollen grains emitting green fluorescence were regarded as viable pollen. Pollen number per flower was counted with a hematocyte arithmometer, and the data is estimated from 30 independent flowers for each line 45.

Chromosome spreading and 4,6-Diamidino-2-phenylindole (DAPI) staining were performed according to Ross et al. 71 with some modifications. Inflorescences collected from rice shoots were fixed in the Carnoy's fixative (methanol: glacial acetic acid = 3:1, v:v) at room temperature at least for 4 h. Then the fixed flowers were washed with 10 mM citrate buffer, pH 4.5, and digested with 0.3% cytohelicase, 0.3% cellulose and 0.3% pectolyase in citrated buffer at 37 °C for 30 min. The digested anthers were washed with citrate buffer and placed in a small drop of 60% acetic acid on a slide, then squashed with a cover slip to release microspore mother cell. After the samples were dried at room temperature, 5 μl DAPI solution (1 μg/ml DAPI, 50% glycerol, 10 mM citrate buffer, pH 4.5) was added onto the slide and then observed under a confocal microscope (Carl Zeiss LSM510 META, Germany). Numbers of the chiasma and bivalent/univalent were counted according to the methods previously described 20, 30, 72.

Quantitative real-time RT-qPCR analysis

Total RNA was extracted using TRizol reagent (Invitrogen) and reverse-transcripted according to the manufacturer's instructions (SuperScript preamplification system; Promega). The primers used for SDS were SDSF4 (5′-AGT GAG GTC GTC GCC ATG GAG TG-3′) and SDSR4 (5′-GGA GGG CCA GAA GGA GAG ATG TTT G-3′). The primers used for RCK were RCKF4 (5′-AGC ATG TGG TCA GCA GGG AAG ATT T-3′) and RCKR4 (5′-GGG CCT CCA GTG AAG CTA AGT TGT TG-3′). Rice ACTIN1 was used as an internal control with primers 5′-GAA CTG GTA TGG TCA AGG CTG-3′ (forward) and 5′-ACA CGG AGC TCG TTG TAG AAG-3′ (reverse).

RT-qPCR was performed on a RotorGene 3000 system (Corbett Research) using a SYBR green detection protocol according to the manufacturer's instructions (SYBR Premix Ex Taq System; TOYOBO). A linear standard curve and threshold cycle number versus log (designated transcript level) were constructed using a series of dilutions of each PCR product; and the levels of the transcript in all unknown samples were determined according to the standard curve. The ACTIN1 was used for normalizing cDNA concentration variations. The experiments were repeated thrice.

Phylogenetic analysis

Homolog sequences in P. trichocarpa, Selaginella moellendorffii and Physcomitrella patens were obtained at JGI website (http://genome.jgi-psf.org), other homologous sequences were obtained from PsiBlast searches at the National center for Biotechnology Information (http://blast.ncbi.nlm.nih.gov/Blast.cgi). The conserved domains of proteins were analyzed at the website http://www.ncbi.nlm.nih.gov/Structure/cdd/wrpsb.cgi. Since all the SDS homologs had both the cyclin_N and cylcin_C domains and all the RCK homologs had the conserved domains including DEXDc, HELICc and SEC63, we used the conserved domains for phylogenetic analyses. Sequence alignments were generated with CLUSTALX 1.83 73, and the alignments between A. thaliana, P. trichocarpa and V. vinifera SDS were adjusted before constructing the tree. Neighbor-joining analyses were performed using MEGA3 74 with the pairwise deletion option, Poisson correction set for distance model and 1 000 bootstrap replicates selected. Maximum likelihood analyses were performed on the PhyML online (http://atgc.lirmm.fr/phyml/) with 100 bootstrap replicates selected 75, 76.